sottens, re buoyancy and DFDR/CVR.
It's not a stupid question; many safety experts and engineers have suggested and even worked on possible designs.
In essence there are some serious problems associated with a design that would work as intended, and with the requisite VERY high reliability. One simple example of a problem with the concept is that, in order for the recorder to not remain trapped under whatever sturcture or parts or debris it might be underneath, it would have to be "jettisoned" somehow, rather than simply "deployed". That in itself leads to a myriad of additional design, engineering, certification, and other issues. As always, any resultant action is a function of cost versus relative benefit. In this example, there are very few instances where recorders have not been retrieved -- regardless of ocean depth or location -- so there is really no incentive to spend the enormous amounts of time and money to change the current specs.